Employment of enhancer elements to drive expression of reporter genes in neurons is a widely used paradigm for tracking axonal projection. For tracking axonal projection of spinal interneurons in vertebrates, germ line-targeted reporter genes yield bilaterally symmetric labelling 151718. Therefore, it is hard to distinguish between the ipsi- and contra-laterally projecting axons. Unilateral electroporation into the chick neural tube provides a useful means to restrict expression of a reporter gene to one side of the central nervous system, and to follow axonal projection on both sides 1920. Mouse enhancer elements are appropriately active in the chick neural tube. Thus, Atoh1, HB9, and HoxA1 enhancer elements drive expression in dI1, motor neurons and floor plate cells, respectively 20212223.

Large-scale transgenic mouse screens of highly conserved non-coding sequences in the human genome have revealed several hundred enhancer elements that target β-galactosidase reporter gene expression to specific developmental structures and cell types in transgenic mice at embryonic day (E)11.5 242526. Two enhancer elements, seemingly expressed in dI neurons, were further analyzed employing in ovo electroporation: #284, located between the Pou3f2 gene and the C6orf167 open reading frame on human chromosome 6; and #169, located between Foxd3 and Atg4c genes on human chromosome 1. The mouse #284 and #169 elements were cloned upstream to Cre recombinase. To verify the specificity to dorsal interneurons, the Enhancers::Cre plasmids were electroporated into stage 16 to 17 chick hemi-tube along with a Cre-dependent mCherry/green fluorescent protein (GFP) plasmid [pCAGG-LoxP-mCherry-LoxP-GFP], which enables the simultaneous detection of the electroporated cells (expressing mCherry) and the enhancer-expressing cells (expressing GFP) (for more details, see Materials and methods). GFP expression is restricted to dorsal neurons (#284; Figure 1A) or the medial lateral neurons (#169, Figure 2A), while mCherry is expressed along the entire ventral/dorsal aspect of the electroporated hemi-tube (Figures 1A and 2A).

To further characterize the cell type specificity of the enhancers, the embryos were co-electroporated along with a Cre-dependent nuclear GFP (nGFP). To determine the identity of reporter-expressing cells, embryos were analyzed at stage 23 to 24 by co-staining with dI-specific antibodies to Lhx2/9 (dI1), Lhx1/5 (dI2, dI4, dI6, V0, V1), Isl1 (dI3), Pax2 (dI4, dI6, V0, V1) and Engrailed1 (V1) (Figure 2I). nGFP expression under the control of #284 is restricted to dI1 neurons, as indicated by co-staining with Lhx2/9 Ab (Figure 1B, C) and the segregation from Lhx1- (Figure 1C) and Isl1-positive neurons (Figure 1B). Of the nGFP-positive neurons, 95.7% (n = 139) are Lhx2/9, and 4.3% (6 of 139) were nGFP-positive but negative to all the above interneuron markers. This minor population may represent progenitors of dI1 that have not upregulated the expression of Lhx2 and Lhx9 yet. The #284 enhancer is herein indicated as EdI1.

Of the neurons that express nGFP under the control of #169, 56% (n = 166) are dI2, as indicated by the co-localization to the dorsal Lhx1/5⁺ (Figure 2B) and Lhx1/5⁺/Pax2^- cells (Figure 2D) and the segregation from Lhx2/9⁺ and Isl1⁺ neurons (Figure 2B, C); 35.5% are V1 neurons as indicated by the localization to the medial Pax⁺/En1⁺/Lhx1/5⁺ neurons (Figure 2D, E). Of the nGFP neurons, 8.5% are presumed to be progenitors of dI2 and V1, since no expression of any cell fate marker was scored. Its proximity to the dI2/V1-specific gene Foxd3 suggests that #169 is a dI2/V1-specific enhancer of Foxd3 (herein indicated as EdI2/V1).

For further focusing on the axonal projection pattern of dI2 neurons, the dI2-specific enhancer element (13G) of the Ngn1 gene was studied in the chick neural tube 27, utilizing the Cre-dependent nGFP system. Expression of nGFP was detected in dorsal/lateral interneurons that express either Lhx2/9 or Lhx1/5 (Additional file 1). Numerous dorsal interneurons located between the ventricular and marginal zones express nGFP. These neurons are presumed to be progenitors of dI1 and dI2 neurons. The expression of Ngn1 in progenitor neurons supports this assumption. The leakage in dI1 neurons while using the Ngn1-13G enhancer 27, versus the entire Ngn1 enhancer 17, suggests that cis elements that are required for repression of expression in dI1 neurons are absent in the 13G enhancer. Hence, in the chick, the Ngn1 13G enhancer is a dI1/2-specific enhancer (herein indicated as Ed1/2).

dI1 neurons give rise to two subpopulations that differ in their cell position, axonal projection and transcription of the Lim-HD proteins: the dI1_comm population, located at the dorsal neural tube and more ventral/medially, which projects axons toward and across the floor plate; and the dI1_ipsi population, located in a ventral/lateral position, which projects axons ipsi-laterally. The division into two subpopulations is also evident in the transcription of the Lim-HD proteins Lhx2 and Lhx9. dI1_comm neurons express Lhx2_high and Lhx9_low, while dI1_ipsi neurons express Lhx9 1628.

The axonal projection pattern of dI1 neurons within the neural tube was studied at E6 utilizing an open-book preparation of electroporated neural tubes. dI1 neurons, labeled with GFP under the control of EdI1 enhancer, project their axons ipsi- and contra-laterally (Figure 3A, B). The neural tubes of ten embryos were analyzed and yielded similar axonal patterns (Table 1). The contra-laterally projecting axons cross the floor plate, turn rostrally and elongate along the floor plate at the ventral funiculus (VF) for a few segments. They are subsequently deflected diagonally and laterally away from the floor plate (Figure 3A, A2, A3). At the lateral funiculus (LF), dI1_comm axons turn rostrally whilst converging to a longitudinal bundle. At the cervical level, a longitudinal fascicule is evidenced only in the LF. All the axons at the contra-lateral VF turn toward the LF. dI1_ipsi axons turn rostrally at the LF (Figure 3A, A4, A5). At sacral levels, few caudally projecting axons are visible (Figure 3A, A1, A3, magenta arrows). The number of caudally versus rostrally projecting axons at the sacral level on the contra-lateral side was scored. 12.5 ± 4.4% (n = 6) of the axons turned caudally. However, it cannot be excluded that beyond E6, more sacral dI1 neurons also project their axons caudally or, alternatively, are eliminated.

The positions of the longitudinal dI1_ipsi and dI1_comm fascicules at the LF along the dorsal/ventral axis appear similar. Hence, it is conceivable that dI1_comm axons from one side of the neural tube, and dI1_ipsi axons from the other side, fasciculate together. To test this hypothesis, GFP or taumyc were expressed in the two halves of the neural tube, respectively (see Materials and methods). Projection of dI1_comm GFP-positive axons toward the dI1_ipsi taumyc axons was inspected (Figure 3C, D, E). dI1_comm axons turned diagonally toward the dI1_ipsi bundle. As they contacted the dI1_ipsi bundle, dI1_comm axons fasciculated with dI1_ipsi and turned rostrally (Figure 3D). Thus, homophilic interaction between dI1_comm and dI1_ipsi may facilitate axonal turning of dI1_comm at the LF.

The Ngn1 enhancer was utilized previously for labelling the axons of dI2 neurons in transgenic mice. Cross-sections and open-book preparation demonstrated that dI2 neurons project their axons toward and across the floor plate 172027. However, the bi-symmetrical expression of the reporter gene precluded detailed mapping of dI2 axonal trajectories.

The axonal cues of dI2 axons were studied utilizing three paradigms: the EdI2/V1 enhancer – V1 neurons project their axons only ipsi-laterally 2930 and, thus, the EdI2/V1 enhancer can be used for studying the contra-lateral projection pattern of dI2 neurons (six embryos; Table 1); the Ed1/2 enhancer – divergence from the dI1 axonal pattern, when employing the dI1/2 enhancer, can be attributed to dI2 neurons (four embryos; Table 1); and molecular intersection of the EdI2/V1 and EdI1/2 enhancers – we have designed a method that enables labelling of neurons that co-express the above enhancers (four embryos; Table 1).

Expression of GFP unilaterally in the chick neural tube under the control of EdI2/V1 revealed that dI2 neurons have two different axonal projection patterns at the contra-lateral side (Figure 4A, Table 1). At the rostral two-thirds of the thoracic level and the brachial and cervical levels, dI2 axons grow toward and across the floor plate. At the contra-lateral side of the floor plate, axons turn rostrally (dI2_rost; Figure 4A, A3). As with dI1_comm axons, dI2_rost axons elongate along the floor plate for a few segments, and subsequently turn laterally and diagonally in the white matter and fasciculate at the LF (Figure 4A). Caudal to the hindlimb, at the lumbar and sacral levels, dI2 axons turn caudally in a mirror-image pattern to the rostrally projecting axons (dI2_caud). Specifically, they grow ventrally toward the floor plate and turn caudally at the contra-lateral side of it (Figure 4A, A1). Then, they turn laterally and form a dI2_caud fascicle at the contra-LF. Along the caudal third of the thoracic level, a mixture of caudally and rostrally projecting axons that form a crisscross pattern at the contra-lateral side is evident (Figure 4A, A2). Similar contra-lateral axonal pathways were observed when employing the dI1/2 enhancer (Figure 4B): rostral (Figure 4B, B3), crisscross (Figure 4B, B2) and caudal (Figure 4B, B1). Since dI1 neurons do not project caudally, the caudal projection that is seen utilizing the dI1/2 enhancer is attributed to dI2 neurons.

The co-expression in dI2 plus V1 neurons or dI2 plus dI1 utilizing the EdI2/V1 and EdI1/2 enhancers, respectively, precludes the identification of ipsi-laterally projecting axons. For labelling dI2 neurons solely, an enhancer intersection technique was adopted. The EdI2/V1 and EdI1/2 enhancers are not exclusive to dI2 neurons; however, their intersection occurs in dI2 neurons. In order to label neurons that co-express EdI2/V1 and EdI1/2 enhancers, we combined the Cre/LoxP and the Gal4/UAS systems. Cre was expressed under the EdI1/2 enhancer, and Gal4 under the EdI2/V1 enhancer. The reporter plasmid contains GFP under a dual control of Gal4 and Cre (UAS-LoxP-STOP-LoxP-GFP; for more details, see Materials and methods). GFP-expressing neurons in which the intersection of EdI1/2 and EdI2/V1 enhancers occurs are Lhx1/5⁺/Pax2^- (100%, n = 22; Figure 2F), Pax2^-/En1^- (100%, n = 18; Figure 2G) and Lhx2/9^-/Isl1^- (100%, n = 29; Figure 2H). Hence, they are dI2 neurons.

The axonal pathways of dI2 neurons at the ipsi- and contra-lateral sides were studied using E6 open book preparations (Figure 5). The axonal patterning of the commissural dI2 axons, labeled exclusively by the intersection technique, is similar to the pattern observed with EdI2/V1 and EdI1/2 enhancers (Figure 4). Namely, dI2_rost axons turn rostrally from the rostral two-thirds of the thoracic level (Figure 5A, A4) and either rostrally or caudally at the caudal third of the thoracic level (Figure 5A, A2, A3), and dI2_caud axons turn caudally from the hindlimb level (Figure 5A, A1). It is difficult to estimate the extent of caudal versus rostral turning at each level due to the axonal abundance of dI2 neurons at the contra-lateral side. However, an inspection of several neural tubes shows that the vast majority of the neurons at the cervical level are dI2_rost (Additional file 2A) and at the sacral level are dI2_caud (Additional file 2B).

At the ipsi-lateral side, few longitudinally projecting axons are seen (Figure 5A, yellow arrow in A5; Additional file 2A). The majority of the axons project circumferentially toward the floor plate (Figure 5A, A6; Additional file 3). No longitudinal tracks are observed at either the VF or the LF. To estimate the ratio between the ipsi- and contra-lateral axonal choice of dI2 neurons, the extent of ipsi-/contra-lateral axons at the cervical level was scored. At this level, no cell bodies were labeled on the ipsi-lateral side. Only in one neural tube (n = 4) were longitudinally projecting axons visible at the cervical ipsi-lateral side (Additional file 3). The ratio between ipsi- to contra-lateral axons is 8.6% in that neural tube. Hence, dI2 neurons have mainly commissural axons that elongate longitudinally either rostrally or caudally, depending on their position along the longitudinal axis.

In motor neurons, reciprocal cross-repression between Isl1 and Lhx1 ensures a sharp boundary between the LMC subpopulations 4. The distinct cell boundaries (Figures 1 and 2) suggest that similar Lim-HD mechanisms may account for dI subdifferentiation. To test whether the Lim-HD code of dI1 and dI2 neurons is maintained through cross-repression, each Lim-HD was expressed uniformly at stage 19 in the chick hemitube (Figure 6). The ratio of neurons co-expressing the ectopic Lim-HD protein and the endogenous Lim-HD protein of the reciprocal dI neurons among the electroporated neurons was measured. In neural tubes electroporated with nGFP, 96% of the electroporated dI1 neurons co-expressed nGFP and Lhx2/9 and 98% of dI2 neurons co-expressed nGFP and Lhx1 (Figure 6C, D). Ectopic expression of Lhx9 resulted in substantial reduction of neurons co-expressing Lhx9 and Lhx1 (Figure 6A). Only 15.76% co-expressed Lhx9 and Lhx1 (Figure 6C). Likewise, Lhx1 affected a comparable decrease in the expression of Lhx9 proteins (Figure 6B). Ectopic Lhx1 resulted in 16.7% of neurons co-expressing Lhx1 together with Lhx9 (Figure 6D).

Cross-repression may arise from a change of cell fate. Thus, ectopic expression of Lhx9 or Lhx1 may determine dI1 and dI2 fate, respectively, which, as a consequence, will lead to down-regulation of the reciprocal Lim-HD protein. However, the competence of repression at a relatively late stage (stage 19) suggests that cross-repression can be mediated in post-mitotic cells without affecting cell fate. The expression of dI1/2 cell fate markers was studied following ectopic expression of Lhx9 and Lhx1. dI2 neurons express Foxd3. The expression of Foxd3 was not up-regulated following Lhx1 ectopic expression (Figure 7A, B). Thus, Lhx1 is not sufficient to impose the complete range of the dI2 cell fate.

The expression of Lhx2 following Lhx9 ectopic expression was used to study dI1 cell fate acquisition. Neurons expressing Lhx9 down-regulate the expression of Lhx2 (Figure 7C, D), suggesting that the segregation to dI1_ipsi neurons, expressing Lhx9, and dI1_comm neurons, expressing Lhx2_high/Lhx9_low, is mediated by cross-repression between Lhx9 and Lhx2. This hypothesis is supported by the reciprocal downregulation of Lhx9 via ectopic expression of Lhx2 (YH and AS, data not shown). Atoh1 is expressed in dI1 progenitor neurons, and its expression is down-regulated in differentiated post-mitotic dI1 neurons. Ectopic expression of Lhx9 causes down-regulation of Atoh1 (Figure 7E, F), suggesting that Lhx9, when up-regulated in post-mitotic dI1 neurons, is a repressor of Atoh1. Barhl1/2 genes are expressed in rodent dI1 neurons 31. However, the orthologous genes are not present in the chick genome. Due to the lack of post-mitotic dI1 markers, the possible repression of dI2 cell fate was studied following Lhx9 ectopic expression. Foxd3 is not repressed in Lhx9-expressing neurons (Figure 7G, H). Lhx9 is thus insufficient for repressing dI2 cell fate. Thus, Lhx9 and Lhx1 cross-repress each other without changing the complete range of cell fate identity. Lhx9 and Lhx1 may nevertheless control some features of differentiated dI neurons. The role of Lhx9 and Lhx1 in axon guidance was studied in the following experiments.

The repression of endogenous Lim-HD following ectopic expression of a reciprocal Lim-HD gene results in replacement of the Lim-HD code. To study the role of the Lim-HD code in the assignment of the axonal projection pattern of dI1 and dI2 neurons, their Lim-HD code was alternated. The following considerations were taken into account in the subsequent ectopic expression experiments. First, to study cell autonomous effects, Lim-HD proteins were expressed specifically in the reciprocal dI neurons utilizing EdI enhancers (Lhx9 in dI2, and Lhx1 in dI1). Second, to follow the axonal trajectories of the manipulated neurons, taumyc or GFP were co-expressed with the ectopic Lim-HD protein from the same plasmid. Third, ectopic expression of Lim-HD may lead to a change in cell properties and, subsequently, to its own down-regulation. For example, Lhx9 expressed in dI2 utilizing the EdI2/V1 enhancer may up-regulate certain dI1 characteristics, ultimately leading to down-regulation of the EdI2/V1 enhancer. The stable Cre/Lox systems were used to stabilize the ectopic expression. Fourth, ectopic expression may result in high, non-physiological levels of exogenous protein levels. The levels of ectopic Lhx9 were compared to the endogenous levels of Lhx2 and Lhx9 (in the non-electroporated side of the neural tube). Utilizing the Cre/Lox system, the exogenous and endogenous levels of Lhx9 were similar (Additional file 4).

Lhx1 controls the projection of LMCl axons to the dorsal limb. LMC neurons that ectopically express Lhx1 settle at the lateral LMC and project their axons to the dorsal limb 4. To test whether Lhx1 may also control the axonal projection of dI2 neurons, it was expressed ectopically in dI1 neurons. The EdI1 enhancer, driving Cre recombinase, was expressed in the neural tube along with an Lhx1/taumyc Cre-conditional plasmid (pCAGG-LoxP-STOP-LoxP-Lhx1-IRES-taumyc; Figure 8; Table 1).

The commissural dI1 and dI2 axonal patterning differ in two aspects (Figures 3, 4 and 5): in the lumbosacral neural tube all dI2_caud axons project caudally, while only the caudal sacral dI1_ipsi axons project caudally; and at the caudal third of the thoracic level, dI2 axons turn either rostrally or caudally, forming a 'crisscross' axonal pattern at the contra-lateral side, while dI1_comm axons turn only rostrally. The consequence of ectopic Lhx1 expression in dI1 neurons (dI1^Lhx1) was studied, focusing on the above features (Figure 8). At the lumbosacral level dI1^Lhx1 axons turn caudally (Figure 8A, A1). At the caudal thoracic level a crisscross pattern of axons turning either rostrally or caudally is evident at the contra-lateral side of the neural tube (Figure 8A, A2). Hence, all the dI2_caud axonal features are assumed by the commissural dI1^Lhx1 neurons (Figure 8B).

Longitudinal axonal tracks of dI1^Lhx1 are present at the ipsi-lateral side (Figure 8A). Hence, Lhx1 does not suppress the ipsi-lateral projection of dI1_ipsi neurons. However, a dI1^Lhx1 fascicule is observed at the ipsi-VF and the ipsi-LF, while dI1_ipsi axons form only an ipsi-LF bundle. The ipsi-VF is a characteristic of V1 axons 29 (Figure 4A), which also express Lhx1. Thus, Lhx1 is sufficient to impose dI2-like and V1-like axonal trajectories to dI1_comm and dI1_ipsi neurons, respectively.

Lhx9 was expressed in dI2 and V1 neurons utilizing the Cre/LoxP systems (EdI2/V1::Cre + pCAGG-LoxP-STOP-LoxP-Lhx9-IRES-GFP; Figure 9; Table 1). The dI2-specific axonal trajectories at the contra-lateral side are described below. Lhx9/GFP-expressing dI2 neurons (dI2^Lhx9) project their axons both rostrally and caudally, forming caudally projecting fascicules at the caudal neural tube level (Figure 9A, A1), followed by a crisscross pattern and rostrally projecting axons (Figure 9A, A2). However, the transition point between caudal, crisscross and rostral projection has been shifted caudally. Wild-type dI2 axons turn caudally from the lumbar level, while dI2^Lhx9 axons turn caudally only in the caudal-sacral level (Figure 9A, A2). The crisscross pattern of dI2 axons is limited to the caudal third of the thoracic level, while dI2^Lhx9 axons either form a crisscross pattern at the rostral-sacral level (four out of six embryos), or do not form it at all (two out of six embryos) (Figure 9B). The rostral turning, which is restricted to levels that lie rostrally to the caudal third of the thoracic level, is expended caudally to the entire thoracic level (five out of six embryos), and even to the lumbar level (three out of six embryos) (Figure 9B). Thus, Lhx9 appears to activate rostral turning at the expense of caudal turning (Figure 9C). A similar axonal patterning was observed when Lhx2 (four embryos; Table 1; Additional file 5) or Lhx2 plus Lhx9 were expressed in dI2 neurons (one embryo; Table 1; Additional file 6). Hence, dI2^Lhx9 axonal trajectories are a mixture of dI1 and dI2 axonal cues. While there are fewer caudally projecting dI2^Lhx9 axons than dI2 axons, caudally projecting axons remain more prevalent than dI1 axons.

Next, we focused on the possible role of Lhx9 on the topographic organization of the longitudinal axonal tracks at the LF. The homophilic fasciculation of the dI1_ipsi + dI1_comm axons at the LF may imply that the longitudinally projecting bundle of dI1 and dI2 axons forms a distinct and specific fascicule. To map the topographic organization of dI1 and dI2 longitudinal tracks, dI1 and dI2 neurons were labeled unilaterally, using the Cre/LoxP method for the commissural dI2 neurons, and the Gal4/UAS method for the dI1 neurons (Figure 10A). Expression of reporter genes in dI1 and dI2 axons reveals that the longitudinal dI2 fascicule at the LF is dorsal to the dI1 fascicle (Figure 10A, C).

The relative position at the LF of dI2^Lhx9 axons was compared to dI1 axons. Taumyc was expressed under the control of the EdI1 enhancer (EdI1::Cre + pCAGG-LoxP-STOP-LoxP-taumyc), together with ectopic Lhx9 in dI2 neurons (EdI2/V1::Gal4 + UAS::Lhx9_UAS::GFP). dI2^Lhx9 axons turn rostrally at the LF, together with dI1_comm axons (Figure 10B, C). The specific bundle of dI2 axons at the LF was not formed, and dI2 and dI1_comm axons intermingled and formed one fascicle as they turned longitudinally at the LF (Figure 10E, Table 1). Thus, Lhx9 may control the homophilic interaction with dI1 axons along their longitudinal projection toward the brain. Alternatively, Lim-HD code may control the position of the LF along the dorsal/ventral axis, where Lhx9 directs a more ventral position than Lhx1, and the mis-expression of Lhx9 in dI2 axons shifts the dorsoventral position at which dI2 axons turn into the longitudinal plane.

The combination of specific enhancers, augmentation of expression levels utilizing the Cre/LoxP and the Gal4/UAS systems, and chick electroporation provide quick and efficient tools for deciphering axonal pathways of a genetically defined group of neurons. The emerging picture is of a complex divergence of axonal cues that arises from dI1 and dI2 subpopulations. dI1 and dI2 give rise to two subpopulations each, defined by the direction of their axonal projections. The simultaneous molecular and spatially restricted labelling of two neuronal populations, dI1 + dI2 and dI1_ipsi+ dI1_comm, underscores the axonal architecture of dI1 and dI2 axons: dI1_ipsi and dI1_comm fasciculate together at the LF; dI1 and dI2 axonal tracks at the LF are segregated.

The axonal pathways of spinal internerons were mapped previously utilizing diI injection 323334. Kadison and Kaprielian described four main axonal projections of decussating axons: intermediate longitudinal commissural (ILC), medial longitudinal commissural (MLC), bifurcating longitudinal commissural (BLC), and forked transverse commissural (FTC) 34. ILC axons travelled rostrally in an arcuate manner, extending into VF regions of the spinal cord before executing a second turn into the longitudinal plane at the LF. The contra-laterally projecting dI1 and dI2 neurons project their axons in an ILC pattern. MLC axons, which extend along the floor plate boundary at the VF for distances greater than 100 mm, BLC axons, which bifurcate to rostral and caudal projections, and FTC axons, which form a trident-shaped or forked projection, were not identified in the current study. The labelling of multiple neurons achieved using the electroporation paradigm in the current study may obscure these projection patterns. A moderate number (10%) of decussated axons was observed to extend in the caudal direction following DiI injection 34. However, our studies point to a larger quantity of caudally projecting neurons. At the sacral level dI1_ipsi and all dI2_caud axons extend caudally. At the lumbar region and the caudal third of thoracic levels, about half of dI2 axons project caudally. Hence, a rostral to caudal stepwise increase in caudal projection is evident. Injection of DiI into neurons at the sacral level may reveal more caudally projecting axons.

The divergence in axonal growth along the ipsi/contra and the caudal/rostral axes may stem from a cell type-specific expression of transcription factors. Namely, dI1_comm and dI2_comm genes could theoretically be expressed in the contra-laterally projecting dI1_comm and dI2 neurons, respectively. A possible candidate for such a mechanism is Lhx2, which is expressed only in dI1_comm neurons. However, gene-targeting experiments of Lhx2 and Lhx9 have shown that a Lim-HD code does not control ipsi- versus contra-lateral axonal projection 16. Similarly, Lhx1 is expressed in the ipsi-only population V1 and the contra-mostly population dI2. Therefore, Lhx1 is probably not implicated in controlling of the contra-lateral projection of dI2 neurons. Alternatively, common dIcomm and dIipsi genes might be expressed in all the dIN_comm and dIN_ipsi neurons, respectively. Transcription factors such as Unc4 and NSCL1, which are expressed in all interneurons 3536 in an overlapping pattern to the commissural-only genes TAG1 and Robo3, are candidates for controlling commissural guidance choice of dIs.

A similar transcriptional mechanism may account for the caudal versus rostral axonal choice. A transcriptional code may discriminate between the longitudinal levels. Hence, the combination of dIcaudal, expressed at the caudal neural tube, and Lhx1 may confer caudal projection. Potential dIcaudal and dIrostral factors may be the Hox proteins. A Hox code determines the rostral/caudal identity of motor neurons, and the combination of Hox and Lim-HD codes determines the subclassification of motor neuron pools 3738. The caudally expressing Hox10 and Hox11 genes may confer caudal turning to the lumbosacral dI2 neurons, while the more rostral Hox, Hox8 and beyond, may confer rostral turning. The caudal thoracic crisscross pattern might be controlled by Hox9+/Hox8- code. The rostral to caudal change in axonal projection, following Lhx1 ectopic expression, may place Lhx1 as a positive activator of the Hoxcaudal genes Hox10 and Hox11. Complementarily, Lhx9 may suppress caudal turning by suppressing Lhx1 and/or Hox10 and Hox11 expression. Alternatively, Lhx9 may suppress Hox10 and Hox11 expression, and Lhx1 may reveal Hox10 and Hox11 activity by suppressing their suppressor, Lhx9.

What are the axonal cues that may govern caudal turning? Axons may turn in different directions due to different axonal cues or differential responsiveness to common cues. The differential cues theory is not supported by our data. At the caudal thoracic levels dI2 axons, at the same rostro/caudal level, turn either rostrally or caudally. The conversion in axonal directionality may be governed cell autonomously by receptors or signalling molecules that convert attraction to repulsion. The rostral turning of commissural neurons along the floor plate is mediated by increasing caudal-to-rostral levels of Wnt proteins 39, which attract axons; and decreasing caudal-to-rostral levels of Shh 40, which repel axons. Caudally turning neurons may express receptors or signalling molecules that convert Wnt attraction to repulsion and/or Shh repulsion to attraction. In vitro assays with caudal dI2 neurons challenged with Wnts and Shh should clarify whether a cell autonomous change in responsiveness governs dorsal and caudal turning.

The emergence of interneuron divisions is marked by a mutual exclusion in the expression profile of bHLH proteins and Lim-HD proteins. Progenitor dI1/2 neurons express Atoh1 and Ngn1/2, respectively 13. Loss and gain of function experiments have demonstrated that these proteins cross repress each other, and are both required and sufficient for the differentiation of dI1/2 neurons 1841. Therefore, in the absence of Atoh1, dI1 neurons fail to differentiate, and are converted to dI2 neurons 18. Lim-HD genes, expressed in the post-mitotic dI1/2 neurons, are probably activated by the bHLH proteins. Our ectopic expression experiments demonstrate that Lim-HD proteins also cross-repress each other in dI1 and dI2 neurons. Thus, the distinct identity of adjacent neurons is guaranteed at the mitotic and post-mitotic stages by cross-repression of bHLH and Lim-HD proteins, respectively. Loss of function experiments have demonstrated that in the absence of Lhx2/9 or Lhx1/5, the fate of dI1 and dI2 neurons is not altered 1642. In the Lhx2/9 double knockout mouse, dI1 cells express dI1-specific genes, and the Lim-HD code is not changed to Lhx1/5. It is conceivable that Atoh1, which acts upstream to Lhx2/9, is repressing Ngn1/2 and thus indirectly prevents the activation of Lhx1/5. It is also possible that bHLH proteins control dI1/2 cell fate by activating Lim-HD proteins and, in addition, in a feed forward mechanism, directly control cell fate. Thus, the elimination of Lim-HD can be compensated for by bHLH proteins. Ectopic Lim-HD proteins may play a dominant role in repressing other Lim-HD proteins and in repressing bHLH protein activity. This assumption is supported by the observation that ectopically expressed Lhx1 suppresses the expression of Atoh1 (YH and OA, unpublished results). The Lim-HD proteins Isl1 and Lhx1 play a similar role in determining the fate of LMC neurons. Retinoic acid induces LMCl neurons by activating Lhx1 and repressing Isl1 expression. In the absence of Lhx1, LMCl neurons differentiate, settle at the lateral LMC column and do not upregulate Isl1 expression. Thus, like bHLH proteins in dI1/2 neurons, retinoic acid is sufficient to confer LMCl identity, probably by bypassing Lhx1 signalling in a feed-forward mechanism 45.

Fertilized white Leghorn chicken eggs were incubated at 38.5 to 39°C. A DNA solution of 5 mg/ml was injected into the lumen of the neural tube at either HH stage 12 to 14 (cytomegalovirus (CMV) enhancer in pCAGG plasmid) or stage 17 to 18 (EdI1 and EdI2/V1 enhancers). For double-sided electroporation, a 1-h interval interceded between the first and second electroporations.

Electroporation was performed using three 50 ms pulses at 25V, applied across the embryo using a 0.5 mm Tungsten wire and a BTX electroporator (ECM 830). Embryos were incubated for 2 to 3 days prior to analysis.

E6 electroporated chick spinal cord tissues were prepared as an open-book preparation by making a longitudinal incision along the roof plate with a sharp tungsten microneedle from the hindbrain down to the tail. The dorsal root ganglia (DRGs) were then separated from the spinal cord, leaving the floor plate intact. The hind and forelimb were marked with charcoal powder, and then the spinal cord was detached from the body and fixed in 4% paraformaldehyde in phosphate-buffered saline for 1 h at room temperature, after which the tissue was spread open to produce flat-mount preparations.

Embryos were fixed overnight at 4°C in 4% paraformaldehyde/0.1 M phosphate buffer, washed twice with phosphate-buffered saline, incubated in 30% sucrose/phosphate-buffered saline for 24 h, and embedded in Optimal Cutting Temperature solution (OCT). Cryostat sections (14 μm) were collected on Superfrost Plus slides and kept at -70°C. Antigen retrieval was used for paraffin sections. Sections were treated with boiled 10 mM citric acid, pH 6, for 10 minutes in the microwave. The following antibodies were used: rabbit polyclonal GFP antibody (Molecular Probes, Eugene, Oregon, USA)), Pax2 (Abcam, Cambridge, MA, USA)), myc (9E10), Isl1 (4D5), Lhx1/5 (4F2), Engrailed, Lhx2/9 (rabbit serum) (all provided by T Jessell, Columbia University, New York, NY, USA). Cy2, RRX and cy5 were used as fluorochromes. Images were taken under a microscope (Axioscope 2; Zeiss) with a digital camera (DP70; Olympus) or confocal microscope (FV1000; Olympus).

The EdI1 enhancer element was amplified by PCR from a genomic mouse DNA utilizing the primers [ATGAGCTCATCCCTTTTTGCTCCCTCAC] and [ATGCTAGCGGTGTTGTGGTTGACAGCAG]. EdI2/V1 was amplified utilizing the primers [ATGAGCTCGCTCTCTCTGCCTACCTCAGC] and [ATGCTAGCAACCTAGTGCCCTTGCACAC]. The enhancers were cloned into 5'SacI/3'NheI sites of the appropriate Cre and Gal4 plasmids 23. The 13G Ngn1 enhancer (EdI1/2) was generated by PCR from a genomic mouse DNA utilizing the primers [ATTGCGGCCGCATCAGGCGCCGGATCACTTTG] and [GATCTAGACCTTCACCATCGTTAACACTGG] and cloned into 5'NotI/3'XbaI sites of the Cre plasmid. Chick Lhx2 and Lhx9 were obtained from Tom Jessell. Chick Lhx1 was obtained from Artur Kania. Chick Foxd3 and Atoh1 were obtained from the chick expressed sequence tag sequencing project 44.

Antisense digoxigenin-labeled probes were prepared by in vitro transcription (Roche, Nutley, NJ, USA)). In situ hybridization was performed on electroporated sections and was combined with immunohistochemistry. Prior to the in situ hybridization, sections were incubated with primary GFP antibody, and then a standard in situ protocol was applied, followed by a secondary fluorescent antibody treatment. Alternatively, adjacent sections were collected on different slides. One set of slides was used for in situ hybridization, and the other for immuno-detection of GFP.